Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Number crunching :
Once in a Blue Moon Challenge
| Author |
Message |
|---|
 Michael Gutierrez Volunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
                 
|
|
Welcome to the Once in a Blue Moon Challenge
The sixth challenge of the 2021 Series will be a 10-day challenge leading up to a relatively rare astronomical event called a blue moon. The challenge will be offered on the PSP-LLR application, beginning 12 August 20:00 UTC and ending 22 August 20:00 UTC.
Once in a blue moon
What if, you can rewrite history?
He asked
For sure
I'll turn it into a fairy tale
He replied
The phrase "blue moon" has actually changed in meaning over the past century. Before the mid-1900s, it simply referred to an impossible event, similar to "when pigs fly". The modern definition of "the second full moon in a calendar month" only became widespread due to a series of misinterpretations in a farmer's almanac, a popular radio show, and funny enough, an edition of the game Trivial Pursuit.
A blue moon happens about once every 2.7 years on average, although the last one was on October 31, 2020. The last PSP prime was found about 3.8 years ago. COINCIDENCE?? (I tried.) Either way, with any luck we can knock out another k-value from the Prime Sierpinski Problem, and prove PSP's to be an (almost) Once in a Blue Moon occurrence.
To participate in the Challenge, please select only the Prime Sierpinski Problem LLR (PSP) project in your PrimeGrid preferences section.
Note on LLR2 tasks: LLR2 has eliminated the need for a full doublecheck task on each workunit, but has replaced it with a short verification task. Expect to receive a few tasks about 1% of normal length.
Application builds are available for Linux 32 and 64 bit, Windows 32 and 64 bit and MacIntel. Intel and recent AMD CPUs with FMA3 capabilities (Haswell or better for Intel, Zen-2 or better for AMD) will have a very large advantage, and Intel CPUs with dual AVX-512 (certain recent Intel Skylake-X and Xeon CPUs) will be the fastest.
Note that LLR is running the latest AVX-512 version of LLR which takes full advantage of the features of these newer CPUs. It's faster than the previous LLR app and draws more power and produces more heat, especially if they're highly overclocked. If you have certain recent Intel Skylake-X and Xeon CPUs, especially if it's overclocked or has overclocked memory, and haven't run the new AVX-512 LLR before, we strongly suggest running it before the challenge while you are monitoring the temperatures.
Multi-threading is supported and IS recommended. (PSP tasks on one CPU core will take 3-4 days on fast/newer computers and 2 weeks+ on slower/older computers.)
Those looking to maximize their computer's performance during this challenge, or when running LLR in general, may find this information useful.
- Your mileage may vary. Before the challenge starts, take some time and experiment and see what works best on your computer.
- If you have a CPU with hyperthreading or SMT, either turn off this feature in the BIOS, or set BOINC to use 50% of the processors.
- If you're using a GPU for other tasks, it may be beneficial to leave hyperthreading on in the BIOS and instead tell BOINC to use 50% of the CPU's. This will allow one of the hyperthreads to service the GPU.
- The new multi-threading system is now live. Click here to set the maximum number of threads. This will allow you to select multi-threading from the project preferences web page. No more app_config.xml. It works like this:
- In the preferences selection, there are selections for "max jobs" and "max cpus", similar to the settings in app_config.
- Unlike app_config, these two settings apply to ALL apps. You can't chose 1 thread for SGS and 4 for SoB. When you change apps, you need to change your multithreading settings if you want to run a different number of threads.
- There will be individual settings for each venue (location).
- This will eliminate the problem of BOINC downloading 1 task for every core.
- The hyperthreading control isn't possible at this time.
- The "max cpus" control will only apply to LLR apps. The "max jobs" control applies to all apps.
- If you want to continue to use app_config.xml for LLR tasks, you need to change it if you want it to work. Please see this message for more information.
- Some people have observed that when using multithreaded LLR, hyperthreading is actually beneficial. We encourage you to experiment and see what works best for you.
Time zone converter:
The World Clock - Time Zone Converter
NOTE: The countdown clock on the front page uses the host computer time. Therefore, if your computer time is off, so will the countdown clock. For precise timing, use the UTC Time in the data section at the very top, above the countdown clock.
Scoring Information
Scores will be kept for individuals and teams. Only tasks issued AFTER 12th August 2021 20:00 UTC and received BEFORE 22th August 2021 20:00 UTC will be considered for credit. We will be using the same scoring method as we currently use for BOINC credits. A quorum of 2 is NOT needed to award Challenge score - i.e. no double checker. Therefore, each returned result will earn a Challenge score. Please note that if the result is eventually declared invalid, the score will be removed.
At the Conclusion of the Challenge
We kindly ask users "moving on" to ABORT their tasks instead of DETACHING, RESETTING, or PAUSING.
ABORTING tasks allows them to be recycled immediately; thus a much faster "clean up" to the end of an LLR Challenge. DETACHING, RESETTING, and PAUSING tasks causes them to remain in limbo until they EXPIRE. Therefore, we must wait until tasks expire to send them out to be completed.
Please consider either completing what's in the queue or ABORTING them. Thank you. :)
About the Prime Sierpinski Project
Wacław Franciszek Sierpiński (14 March 1882 - 21 October 1969), a Polish mathematician, was known for outstanding contributions to set theory, number theory, theory of functions and topology. It is in number theory where we find the Sierpinski problem.
Basically, the Sierpinski problem is "What is the smallest Sierpinski number" and the prime Sierpinski problem is "What is the smallest 'prime' Sierpinski number?"
First we look at Proth numbers (named after the French mathematician François Proth). A Proth number is a number of the form k*2^n+1 where k is odd, n is a positive integer, and 2^n>k.
A Sierpinski number is an odd k such that the Proth number k*2^n+1 is not prime for all n. For example, 3 is not a Sierpinski number because n=2 produces a prime number (3*2^2+1=13). In 1962, John Selfridge proved that 78,557 is a Sierpinski number...meaning he showed that for all n, 78557*2^n+1 was not prime.
Most number theorists believe that 78,557 is the smallest Sierpinski number, but it hasn't yet been proven. In order to prove it, it has to be shown that every single k less than 78,557 is not a Sierpinski number, and to do that, some n must be found that makes k*2^n+1 prime.
The smallest proven 'prime' Sierpinski number is 271,129. In order to prove it, it has to be shown that every single 'prime' k less than 271,129 is not a Sierpinski number, and to do that, some n must be found that makes k*2^n+1 prime.
Previously, PrimeGrid was working in cooperation with Seventeen or Bust on the Sierpinski problem and working with the Prime Sierpinski Project on the 'prime' Sierpinski problem. Although both Seventeen or Bust and the Prime Sierpinski Project have ceased operations, PrimeGrid continues the search independently to solve both conjectures.
The following k's remain for each project:
Sierpinski problem (SoB) Prime Sierpinski problem (PSP) Extended Sierpinski Problem (ESP)
21181 22699* 91549
22699 67607* 131179
24737 79309 163187
55459 79817 200749
67607 152267 202705
156511 209611
222113 227723
225931 229673
237019 238411
being tested by Seventeen or Bust
Additional Information
For more information about Sierpinski, Sierpinski number, and the Sierpinsk problem, please see these resources:
What is LLR?
The Lucas-Lehmer-Riesel (LLR) test is a primality test for numbers of the form N = k*2^n − 1, with 2^n > k. Also, LLR is a program developed by Jean Penne that can run the LLR-tests. It includes the Proth test to perform +1 tests and PRP to test non base 2 numbers. See also:
What is LLR2?
LLR2 is an improvement to the LLR application developed by our very own Pavel Atnashev and stream. It utilizes Gerbicz checks to enable the Fast DoubleCheck feature, which will nearly double the speed of PrimeGrid's progress on the projects it's applied to. For more information, see this forum post.
____________
Time flies like an arrow. Fruit flies like a banana.
| |
|
KEPSend message
Joined: 10 Aug 05
Posts: 293
ID: 110
Credit: 9,443,250
RAC: 3,891
      
|
|
I want to participate in this challenge, the first in about a year. Unfortunantly I have lost track of the thread contatining the information about the current FFT length of each task in each LLR subproject - so can anyone tell me the current minFFT and maxFFT for the Prime Sierpinski Problem?
I'm going to use the information about minFFT and maxFFT to run as most task as possible on the 12 core Xeon. Something tells me that running 2 tasks with 6 cores each may be the best option, but dependant on the FFT size, 3 task using 4 cores each might also be possible :)
A wish could be the very usefull information about the current FFT length are put in a post in an easy to find place, such that no one has to ask this question again later on. It could possibly be in the Number crunching thread :)
Have a nice one everyone :) | |
|
|
|
I want to participate in this challenge, the first in about a year. Unfortunantly I have lost track of the thread contatining the information about the current FFT length of each task in each LLR subproject - so can anyone tell me the current minFFT and maxFFT for the Prime Sierpinski Problem?
I'm going to use the information about minFFT and maxFFT to run as most task as possible on the 12 core Xeon. Something tells me that running 2 tasks with 6 cores each may be the best option, but dependant on the FFT size, 3 task using 4 cores each might also be possible :)
A wish could be the very usefull information about the current FFT length are put in a post in an easy to find place, such that no one has to ask this question again later on. It could possibly be in the Number crunching thread :)
Have a nice one everyone :)
2400K-2880K afaik.
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
 | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
                       
|
|
I've been running PSP for days now.
I can only find FFT sizes of 2240k, 2304k, 2400k and 2520k.
This may change during the challenge to higher FFT. | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
I'm going to use the information about minFFT and maxFFT to run as most task as possible on the 12 core Xeon. Something tells me that running 2 tasks with 6 cores each may be the best option, but dependant on the FFT size, 3 task using 4 cores each might also be possible :)
I am guessing your xeon is similar to my X intel CPUs.
You will probably be best running only 1 task at once.
Unless you don't have enough cache (L2 + L3 for xeon?) but even then 1 task may still be best.
With L2 + L3 - I think - and I just made this up - L2 + 1/2 L3 - is going to work.
As an example, running all 16 cores on a 9960X, which then has 16Mb L2 cache - I know mid 20s will be fine, but high 20s and low 30s not. L3 cache on a 9960X is 22 Mb. 16 + 11 = 27 which is about right for as much 'working in cache' as you could hope for.
Edit: Mind you I run 2 x CUL or 2 x WO0 which uses about 32Mb and it only slows it down a bit - to about the speed of a 9980XE. | |
|
|
|
|
2520? Shouldnt it be 2560.... But yes in that ballpark
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
2520? Shouldnt it be 2560.... But yes in that ballpark
Definitely 2520k. I remember someone saying FFT may be different for AVX512 - so these numbers may apply to few people. Not that there will be much difference, if there is a difference. | |
|
KEPSend message
Joined: 10 Aug 05
Posts: 293
ID: 110
Credit: 9,443,250
RAC: 3,891
|
|
Dear guys.
It appears that the suggestion of running 1 task is the most efficient.
Things sure has changed since last I ran Prime Sierpinski Problem tasks.
Just to be sure, I'll do another testing to see if the 30MB L3 cache is exhausted such that I get a memory bottleneck when running 2 tests using 6 cores. If running 1 test at a time using all 12 cores seems to be the most efficient, the Xeon itself will complete approximately 22 firsttime checks during the Once in a Blue Moon challenge :)
I'll get back with my data, once I have a clear comparison. Thanks for all your feedback.
The test that I ran using 12 cores a test, is using FFT=2560K.
Take care. | |
|
KEPSend message
Joined: 10 Aug 05
Posts: 293
ID: 110
Credit: 9,443,250
RAC: 3,891
|
|
Data is in. It is at least 2.48% faster to use 12 cores per test on the Xeon :)
Take care and thanks for your feedback.
See you at the challenge :) | |
|
|
|
If you have a CPU with hyperthreading or SMT, either turn off this feature in the BIOS, or set BOINC to use 50% of the processors
Could you explain why's that recommended? | |
|
|
|
|
HyperThreading/SMT performance improvements are mainly in the ability to do work on one thread while another thread is waiting for data to be fetched. This is accomplished by some extras registers that store the state of each thread such that it's very fast switch between threads.
For example, if one thread needs to fetch something from memory, another thread could run computations, and when that thread was finished computing the first thread probably has the data it needs and can continue computing. This leads to a noticable improvement in most applications that have a mix of computation and data manipulation.
However, PrimeGrid is almost entirely computation with very little data fetching. If you were to run PrimeGrid on two threads sharing the same CPU core, each thread would be "fighting" for computation units in the core (there's only a fixed number per core, it's the same with or without SMT). This would lead to extra thread switching. While very fast, this non-zero time thread switch does "waste" time in the sense that it's not used for meaningful computation.
Additionally, any time you have concurrent algorithms in programming, there will be some penalty for synchronizing the data. If two threads write to the same memory location at the same time, the data in that location likely gets corrupted. There are typically locks where one thread writes and the others wait their turn in line to write next. This slowdown can be seen by running a typical multicore program on 2 cores, and noticing it will take slightly longer than half the time of running it on one core (when you might naively expect 2 cores to be exactly twice as far as 1). The extra time is generally all of the data synchronization. With hyperthreading/SMT, you're essentially doubling the amount of worker threads, which leads to double the time wasted on data synchronization.
This is a pretty simplified overview. There are a lot more layers of computer that impact how everything works, but these are some of the pieces that I'm somewhat familiar with. | |
|
|
|
I'm going to use the information about minFFT and maxFFT to run as most task as possible on the 12 core Xeon. Something tells me that running 2 tasks with 6 cores each may be the best option, but dependant on the FFT size, 3 task using 4 cores each might also be possible :)
I am guessing your xeon is similar to my X intel CPUs.
You will probably be best running only 1 task at once.
Unless you don't have enough cache (L2 + L3 for xeon?) but even then 1 task may still be best.
With L2 + L3 - I think - and I just made this up - L2 + 1/2 L3 - is going to work.
As an example, running all 16 cores on a 9960X, which then has 16Mb L2 cache - I know mid 20s will be fine, but high 20s and low 30s not. L3 cache on a 9960X is 22 Mb. 16 + 11 = 27 which is about right for as much 'working in cache' as you could hope for.
Edit: Mind you I run 2 x CUL or 2 x WO0 which uses about 32Mb and it only slows it down a bit - to about the speed of a 9980XE.
I'm guessing that's the power of AVX-512 right there.
My testing came up with the following -
AMD
3900X running 1 task x 12 cores = 41,000 sec run time.
3900X running 2 tasks x 12 cores = 66,000 sec run time.
1950X running 1 task x 16 cores = 46,000 sec run time.
1950X running 2 tasks x 16 cores = 87,000 sec run time.
3950X running 1 task x 16 cores = 52,000 sec run time.
3950X running 2 tasks x 16 cores = 74,000 sec run time.
I've yet to try dividing again but I have a feeling running 2 wu's will be best. Just a hunch. I'll test with the 3900X
____________
 | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
            
|
|
You can use pfgw to find the FFT size of any subproject:
>pfgw64 -q"237019*2^24445928+1" -V
PFGW Version 4.0.1.64BIT.20191203.Win_Dev [GWNUM 29.8]
Special modular reduction using all-complex AVX FFT length 2560K, Pass1=640, Pass2=4K, clm=4 on 237019*2^24445928+1
>pfgw64 -q"79309*2^24445928+1" -V
PFGW Version 4.0.1.64BIT.20191203.Win_Dev [GWNUM 29.8]
Special modular reduction using all-complex AVX FFT length 2304K, Pass1=384, Pass2=6K, clm=1 on 79309*2^24445928+1
The size of k has a big impact on FFT size. To find the n currently worked on, check the status pages.
The FFT size for your system might be slightly different, depending on which instruction sets are available.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
|
|
|
Dear co-crunchers,
Being a newbie, I'd like to clear one issue. If I participate in this challenge (for CPU), can I left one independent GPU-task working (e.g. GFN16)? In this case, will my activity be counted against PSP or I will receive zero points? | |
|
|
|
Dear co-crunchers,
Being a newbie, I'd like to clear one issue. If I participate in this challenge (for CPU), can I left one independent GPU-task working (e.g. GFN16)? In this case, will my activity be counted against PSP or I will receive zero points?
Points are based on credit received in the PSP project, so you are fine to run other projects in parallel.
____________
 | |
|
|
|
You can use pfgw to find the FFT size of any subproject:
Will PFGW always select the same FFT size as LLR/LLR2? /JeppeSN | |
|
|
|
HyperThreading/SMT performance improvements are mainly in the ability to do work on one thread while another thread is waiting for data to be fetched. This is accomplished by some extras registers that store the state of each thread such that it's very fast switch between threads.
(...)
However, PrimeGrid is almost entirely computation with very little data fetching. If you were to run PrimeGrid on two threads sharing the same CPU core, each thread would be "fighting" for computation units in the core (there's only a fixed number per core, it's the same with or without SMT)(...)
Thanks mate, that's more than helpful. | |
|
KEPSend message
Joined: 10 Aug 05
Posts: 293
ID: 110
Credit: 9,443,250
RAC: 3,891
|
You can use pfgw to find the FFT size of any subproject:
Will PFGW always select the same FFT size as LLR/LLR2? /JeppeSN
If the GWnum version of LLR/LLR2 and PFGW is the same, the answer is YES!
However, from time to time, George updates the GWnum library and in the case where you close in on a FFT crossover, then the FFT length can start to vary.
In other words, even if the GWnum version is not the same, then in most of cases, the FFT selected will be the same. | |
|
|
|
|
I've given up on the challenge before it even starts.
Even my fastest computer won't be able to complete even one work unit in the 10 day contest.
____________
The biggest threat to public safety and security is not terrorism, it is Government abuse of authority.
Bitcoin Donations: 1Le52kWoLz42fjfappoBmyg73oyvejKBR3 | |
|
KEPSend message
Joined: 10 Aug 05
Posts: 293
ID: 110
Credit: 9,443,250
RAC: 3,891
|
|
Here is some fun numbers, each one can use their own way:
1. We have below n=50M about 355,472 k/n pairs remaining.
2. Everytime we complete 1000 k/n pairs, we increase maximum n with about 71,023
3. Everytime we complete 14,080 k/n pairs, we increase maximum n with about 1,000,000
4. If we complete 3554 k/n pairs each day, we will clear 10% of work remaining by the end of the challenge
5. If we complete 8,885 k/n pairs each day, we will clear 25% of work remaining by the end of the challenge
6. If we find a prime during this challenge, we will save from 22,420 to 104,827 k/n pairs to test, dependant of the weight of the k
Have fun everyone :) | |
|
|
|
I've given up on the challenge before it even starts.
Even my fastest computer won't be able to complete even one work unit in the 10 day contest.
After all, you're using Core2 Duos... you need an upgrade
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
|
|
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
____________
| |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2926
ID: 130544
Credit: 1,527,307,352
RAC: 3,772,924
                     
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
There are ways & means. | |
|
|
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
There are ways & means.
Yes...if only I could access BOINCTASKS from my phone!
She'll be right. I'm not a serious contender with these comps anyway. Its a bit of fun :)
____________
| |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
What time do you start work?
Challenge starts at 6am (Melbourne, Sydney, Brisbane) | |
|
|
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
Hah! We are just starting a fresh heat wave, today. Highs starting at 95F/35C, so I'll be delaying my start for a few days.
(In my area, this is quite uncommon, especially for a second time in a year, and we have one of the lowest AC installation rates in the country)
____________
Eating more cheese on Thursdays. | |
|
|
|
|
3:00 P.M. in Chicago, Illinois, USA.
O.K. for those such as myself who are not working. | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
|
In waiting for the challenge to start I had switched to SGS and thought that with nearly 30,000 tasks a second prime was due. And I actually found one and despite a 35% 1st rate, I finished first.
6348086090157*2^1290000-1
It's not much, but it's honest work.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
3:00 P.M. in Chicago, Illinois, USA.
O.K. for those such as myself who are not working.
I hope all is well in the windy city | |
|
|
|
I've just realised the start time SUCKS for this comp!! LOL
I'll be at work! I'll miss out on at least 6hrs of crunching :(
That's a better alternative than starting early though.
What time do you start work?
Challenge starts at 6am (Melbourne, Sydney, Brisbane)
Start at 6am mate.
I'm up at 4:30am for work every day.
____________
| |
|
|
|
|
Greetings friends.
How will my i5 520M handle these heavy tasks? I hope I can get some extra points.
Much encouragement to all.
____________
| |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
Greetings friends.
How will my i5 520M handle these heavy tasks? I hope I can get some extra points.
Much encouragement to all.
Greeting and salutations!
With your CPU, do 1 task at once with all cores.
May the force be with you.
| |
|
|
|
|
Dog Days of Summer.
Yesterday was 94 degrees and 94% Humidity.
Plus 5 inches of rain this week.
Will stay in the basement where it is cool, with my "children" and crunch. | |
|
|
|
Dog Days of Summer.
Yesterday was 94 degrees and 94% Humidity.
Plus 5 inches of rain this week.
Will stay in the basement where it is cool, with my "children" and crunch.
93 here today, and will keep on for 3 days till sunday/monday with no rain... :(
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
|
|
In waiting for the challenge to start I had switched to SGS and thought that with nearly 30,000 tasks a second prime was due. And I actually found one and despite a 35% 1st rate, I finished first.
6348086090157*2^1290000-1
It's not much, but it's honest work.
Congratulations. /JeppeSN | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
Dog Days of Summer.
Yesterday was 94 degrees and 94% Humidity.
Plus 5 inches of rain this week.
Will stay in the basement where it is cool, with my "children" and crunch.
93 here today, and will keep on for 3 days till sunday/monday with no rain... :(
Meanwhile in Germany we had our four days of summer as it seems this weekend. After the hot last few years I forgot how an average north/west german summer is like.
I guess many of you will envy me, maybe it's just that one always wants what one doesn't have. ;)
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
|
|
Dog Days of Summer.
Yesterday was 94 degrees and 94% Humidity.
Plus 5 inches of rain this week.
Will stay in the basement where it is cool, with my "children" and crunch.
93 here today, and will keep on for 3 days till sunday/monday with no rain... :(
Meanwhile in Germany we had our four days of summer as it seems this weekend. After the hot last few years I forgot how an average north/west german summer is like.
I guess many of you will envy me, maybe it's just that one always wants what one doesn't have. ;)
Ahhhh usually it gets to around 30C and dips back in mid-August to somewhere like 70F...this year is especially hot.
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
|
|
|
Hallo everyone,
this time I can not take part on the challenge. My new Xeon 40 core 80 threads is not ready. I m still waiting on 64GB DDR4 ECC 3200MHz Server RAM from CN. Furthermore I need still some of the newest GPUs from AMD or NVIDIA for a reasonable price.
Meanwhile I use my old GTX570 to run 4 PPS Sieve and I reduced the CPU via BOINC on 40 Cores.
http://www.primegrid.com/show_host_detail.php?hostid=1084626
____________
| |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
One day down, nine to go! Stats so far:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-13 19:20:33 UTC)
9355 tasks have been sent out. [CPU/GPU/anonymous_platform: 9355 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
2400 (26%) were aborted. [2400 (26%) / 0 (0%) / 0 (0%)]
23 (0%) came back with some kind of an error. [23 (0%) / 0 (0%) / 0 (0%)]
581 (6%) have returned a successful result. [581 (6%) / 0 (0%) / 0 (0%)]
6351 (68%) are still in progress. [6351 (68%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
382 (66%) are pending validation. [382 (66%) / 0 (0%) / 0 (0%)]
199 (34%) have been successfully validated. [199 (34%) / 0 (0%) / 0 (0%)]
0 (0%) were invalid. [0 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25086363. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 1.96% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 2 updates:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-15 01:28:27 UTC)
16638 tasks have been sent out. [CPU/GPU/anonymous_platform: 16638 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
4540 (27%) were aborted. [4540 (27%) / 0 (0%) / 0 (0%)]
227 (1%) came back with some kind of an error. [227 (1%) / 0 (0%) / 0 (0%)]
2381 (14%) have returned a successful result. [2381 (14%) / 0 (0%) / 0 (0%)]
9490 (57%) are still in progress. [9490 (57%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1054 (44%) are pending validation. [1054 (44%) / 0 (0%) / 0 (0%)]
1327 (56%) have been successfully validated. [1327 (56%) / 0 (0%) / 0 (0%)]
0 (0%) were invalid. [0 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25442643. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 3.40% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
Hi! I'm taking part on this challenge.
The first task I sent lasted for 10 hours approx.
Now the estimates are for more that 1.5 days!
Is this normal? Are the WU progressively more complex to finish?
Thanks!
____________
| |
|
|
|
Hi! I'm taking part on this challenge.
The first task I sent lasted for 10 hours approx.
Now the estimates are for more that 1.5 days!
Is this normal? Are the WU progressively more complex to finish?
Thanks!
The work units will not grow so fast.
I can see your PSP work units on https://www.primegrid.com/results.php?hostid=1080697&show_names=1&appid=8.
It looks like you got several work units simultaneously at 04:24 and 04:37 today. If you change a setting on how many tasks to run in parallel (degree of multithreading), that can influence the running times.
Also, sometimes the estimated times can be wrong, for some reason.
Ideally, the credit should be a reasonable measure for the size of the task. I see your completed tasks in the challenge have had credits:
41,796.66
41,922.24
(proof task 328.83)
42,358.10
This illustrates that typically the size will grow, but only slowly.
/JeppeSN | |
|
|
|
|
Thanks JeppeSN!
Prior to the start of the challenge I should have tested what works best in my case: one task at a time using 16 cores, or two tasks running in parallel using 8 cores each. (This is what I'm running now.)
What do you think it's better?
I will let them finish to compare the real running times with the previous tasks.
Salut! | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
Thanks JeppeSN!
Prior to the start of the challenge I should have tested what works best in my case: one task at a time using 16 cores, or two tasks running in parallel using 8 cores each. (This is what I'm running now.)
What do you think it's better?
I will let them finish to compare the real running times with the previous tasks.
Salut!
I think your CPU only has 8 cores.
I can't find info on CCXs.
I am guessing the CPU has 2 x 4 core.
But 32 Mb total L3 cache.
Most likely best option is running only 1 x 8 core task (no hyper-threading)
2nd option is 2 x 4 core tasks but I will guess any gains not crossing CCXs will be lost by going beyond L3 cache. | |
|
GDBSend message
Joined: 15 Nov 11
Posts: 261
ID: 119185
Credit: 2,859,217,799
RAC: 357,194
          
|
|
It would be very nice if there was a PG project switch to send ONLY DC tasks for LLR subprojects. Without such a switch, I will be forced to download a 10 day queue of PSP tasks, and then abort any that I can't finish in the time remaining.
____________
 | |
|
|
|
I think your CPU only has 8 cores.
I can't find info on CCXs.
I am guessing the CPU has 2 x 4 core.
But 32 Mb total L3 cache.
Most likely best option is running only 1 x 8 core task (no hyper-threading)
2nd option is 2 x 4 core tasks but I will guess any gains not crossing CCXs will be lost by going beyond L3 cache.
You're right, 8 cores with 2 threads/core and 32 MiB L3 cache:
# lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 43 bits physical, 48 bits virtual
CPU(s): 16
On-line CPU(s) list: 0-15
Thread(s) per core: 2
Core(s) per socket: 8
Socket(s): 1
NUMA node(s): 1
Vendor ID: AuthenticAMD
CPU family: 23
Model: 113
Model name: AMD Ryzen 7 3700X 8-Core Processor
Stepping: 0
Frequency boost: enabled
CPU MHz: 3993.834
CPU max MHz: 5224,2178
CPU min MHz: 2200,0000
BogoMIPS: 7186.35
Virtualization: AMD-V
L1d cache: 256 KiB
L1i cache: 256 KiB
L2 cache: 4 MiB
L3 cache: 32 MiB
NUMA node0 CPU(s): 0-15
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Full AMD retpoline, IBPB conditional, STIBP always-on, RSB filling
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm cons
tant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdr
and lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfct
r_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate sme ssbd mba sev ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sh
a_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd arat npt lbrv svm_lock n
rip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip rdpid overflow_recov succor smca
I admit I must study well all those concepts concerning CPU architecture to understand what's happening under the hood...
Following your recommendation, I would like to disable SMT with
# echo off > /sys/devices/system/cpu/smt/control
Can this be done while running tasks or something could get corrupted? Should I suspend, turn SMT off and resume? Maybe it's better to let the tasks end and turn SMT off before a new task starts?
I know, so many newbie questions... :_)
Currently, this is my situation:
Thanks a lot for your guidance! | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
|
If you are running 2 x 8 thread tasks - when you change the setting from 16 threads to 8 - it will leave one of those tasks waiting to run and will finish it when the other is finished. There shouldn't be an issue changing the number of threads to run during a task. Only if you made the setting less than 8, the 8 thread task shouldn't run (I have had weird things happen)
In Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 50% of the processors
Or you could turn off SMT in the bios.
Then in Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 100% of the processors | |
|
|
|
|
Your CPU is running just fine and even with SMT/HT around 100% faster (?) than my Ryzen 7 3700X wit SMT/HT deactivated - so you should be fine. Albeit you are running Linux and me Windows 10, i am running 2 x 4 Threads and you are running 2 x 8 Threads.
The WU-scheduling "should" (with a large grain of salt) be handled via the process scheduler in Linux - it is HT/CCX-aware to a better degree than you could be with manually pinning WUs to cores.
Core-topology is normaly stored somewhere around /sys/devices/system/cpu/cpu${number}/topology/core_siblings_list
numactl --hardware should give you an insight if the kernel thinks some "cores" (more like domains) have fancy relationship towards each other (but your lscpu shows one numa-node, so that is out of the question).
The MT-settings are a bit coarse:
I got 1 4-Core, 2 8-Core, 1 2x8-Core and 1 12-Core (not running at the moment), so 4 Threads is a safe bet for me.
I clocked several Xeon (E5 v1 to v3, E7, Silver/Bronze/Gold) with 321-llr and i think 3k-FFTs but i couldn't fully get a clear speedup recomendation regarding cache-sizes.
For disabling SMT/HT i would go the EFI/BIOS-route, or adding nosmt as a grub2 kernelopts-parameter. | |
|
|
|
If you are running 2 x 8 thread tasks - when you change the setting from 16 threads to 8 - it will leave one of those tasks waiting to run and will finish it when the other is finished. There shouldn't be an issue changing the number of threads to run during a task. Only if you made the setting less than 8, the 8 thread task shouldn't run (I have had weird things happen)
In Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 50% of the processors
Or you could turn off SMT in the bios.
Then in Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 100% of the processors
I fully agree with Nick, if you have already disabled SMT in BIOS, then you should just use 100% of the processors in the BOINC manager. You should also set on PG Prefs page what nick said:
"use at most 8 processors".
Good luck crunching!
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
|
|
In Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 50% of the processors
Or you could turn off SMT in the bios.
Then in Computing Preferences:
On multiprocessors, use at most 8 processors
On multiprocessors, use at most 100% of the processors
You mean this?
 | |
|
|
|
You mean this?
Oh forget it, I already found it! | |
|
|
|
|
Hi guys, I am getting alarm messages on completed tasks.
https://www.primegrid.com/results.php?hostid=1084312&appid=8
What could I be doing wrong?
____________
| |
|
|
|
Hi guys, I am getting alarm messages on completed tasks.
https://www.primegrid.com/results.php?hostid=1084312&appid=8
What could I be doing wrong?
Most likely there is something going wrong with your hardware. It appears that all of your tasks are having issues, but only certain ones trigger the warning flag. The good news is that they are at least completing, but how long that will last is hard to tell.
Some things to check, remembering your last thread on cooling: How are your system temperatures? Overheating causes errors, could be CPU, could be memory, could be something else. Great place to start making sure there is lots of air moving through the system and that the dust is blown out, as well as checking the CPU temps.
You did turn off AMD Boost/PBO in the BIOS, last time, right? Are the voltages set to "auto" or did you enter values?
Is your system otherwise stable? Do you get any BSODs/crashes in Windows? Funky behavior?
____________
Eating more cheese on Thursdays. | |
|
|
|
|
I am coming home and now I will be able to check the BIOS.
I set the cpu speed to 3.6mhz voltaje auto, maybe that is what is giving problem ?
(the temperatures in my city in summer reach 40 degrees Celsius)
____________
| |
|
|
|
I am coming home and now I will be able to check the BIOS.
I set the cpu speed to 3.6mhz voltaje auto, maybe that is what is giving problem ?
(the temperatures in my city in summer reach 40 degrees Celsius)
Heat will be an issue sometimes too.
____________
SHSIDElectronicsGroup@outlook.com
waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1
| |
|
|
|
|
I'm home now.
The CPU temperature is stable, it's running at 3,750Mhz and this helps with that. (I don't know if this is a good idea)
Precission bost is disabled.
Is there any way to see if the task is running correctly if I make changes to the bios?
HWinfo shows speed drops as shown in the picture and I don't quite know why. could it be that?
Thanks for your help.
Translated with www.DeepL.com/Translator (free version)
____________
| |
|
|
|
|
Core temperatur is just fine and the frequency is only 100 MHz below the lowest max frequency.
I could not see the Package Power or Core Power and would presume those are the restricting factors.
My E5-2690 also clocks down from its 3.3 GHz TurboBoost to around 3.1/3.2 since it draws 145 W with the former and down to the specced 135 W with the later. (~85 °C, borderline ok ;) water cooling would help i think, but i would not sink more money into this rig).
On your front it could be a faulty memory module or (a rare) faulty CPU(-part). | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 3 Updates:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-16 02:20:55 UTC)
20217 tasks have been sent out. [CPU/GPU/anonymous_platform: 20217 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
5773 (29%) were aborted. [5773 (29%) / 0 (0%) / 0 (0%)]
318 (2%) came back with some kind of an error. [318 (2%) / 0 (0%) / 0 (0%)]
4159 (21%) have returned a successful result. [4159 (21%) / 0 (0%) / 0 (0%)]
9967 (49%) are still in progress. [9968 (49%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1596 (38%) are pending validation. [1596 (38%) / 0 (0%) / 0 (0%)]
2563 (62%) have been successfully validated. [2563 (62%) / 0 (0%) / 0 (0%)]
0 (0%) were invalid. [0 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25603821. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 4.06% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
I have modified the RAM working frequency, just in case. I am waiting for the tasks to finish.
I'm having bad luck....
____________
| |
|
|
|
It would be very nice if there was a PG project switch to send ONLY DC tasks for LLR subprojects. Without such a switch, I will be forced to download a 10 day queue of PSP tasks, and then abort any that I can't finish in the time remaining.
Try using the top setting here:
and setting it for each sub-project so you only get X number of tasks at a time. If you are asking for a 'master switch' for every sub-project that would be handled thru all the different venues that PrimeGrid offers us. You could set each of the different venues for a different sub-project and then move your pc's from venue to venue as you see fit. | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 920
ID: 301928
Credit: 513,286,703
RAC: 5,278
               
|
It would be very nice if there was a PG project switch to send ONLY DC tasks for LLR subprojects. Without such a switch, I will be forced to download a 10 day queue of PSP tasks, and then abort any that I can't finish in the time remaining.
If was discussed few times already and it's just not possible. To generate stable flow of DC tasks for just one computer, 64 slave computers working on main tasks will be required.
| |
|
GDBSend message
Joined: 15 Nov 11
Posts: 261
ID: 119185
Credit: 2,859,217,799
RAC: 357,194
|
It would be very nice if there was a PG project switch to send ONLY DC tasks for LLR subprojects. Without such a switch, I will be forced to download a 10 day queue of PSP tasks, and then abort any that I can't finish in the time remaining.
If was discussed few times already and it's just not possible. To generate stable flow of DC tasks for just one computer, 64 slave computers working on main tasks will be required.
A DC ONLY switch would normally be used only by users during challenges when long tasks won't finish
before the challenge ends. As part of this, at the PG admin level, a switch is needed to set a subproject
into CHALLENGE MODE. In CHALLENGE MODE, DC tasks are ONLY sent to users that explicitly request
them via DC ONLY switch.
During a challenge you may have time to finish a last LONG task. But before you get your last LONG task,
one or more DC tasks are downloaded and run. Now your last LONG task will finish after the challenge
ends and you miss getting credit for it.
With CHALLENGE MODE set by PG admin for a subtask, DC tasks will accumulate until users set DC ONLY
near the end of the challenge and drain the queue. There will be little or no DC cleanup at the end of the challenge.
____________
| |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2926
ID: 130544
Credit: 1,527,307,352
RAC: 3,772,924
|
|
Sounds like extra work for everyone to me | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13648
ID: 53948
Credit: 285,351,178
RAC: 47,109
       
|
It would be very nice if there was a PG project switch to send ONLY DC tasks for LLR subprojects. Without such a switch, I will be forced to download a 10 day queue of PSP tasks, and then abort any that I can't finish in the time remaining.
If was discussed few times already and it's just not possible. To generate stable flow of DC tasks for just one computer, 64 slave computers working on main tasks will be required.
A DC ONLY switch ...
Let me help out with the word you're having trouble with...
impossible [ im-pos-uh-buhl ]
adjective
- not possible; unable to be, exist, happen, etc.
- unable to be done, performed, effected, etc.:
an impossible assignment.
- incapable of being true, as a rumor.
- not to be done, endured, etc., with any degree of reason or propriety:
an impossible situation.
- utterly impracticable:
an impossible plan.
- hopelessly unsuitable, difficult, or objectionable.
(From dictionary.com.)
____________
My lucky number is 75898524288+1
| |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 4:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-17 02:06:42 UTC)
23473 tasks have been sent out. [CPU/GPU/anonymous_platform: 23473 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
8247 (35%) were aborted. [8247 (35%) / 0 (0%) / 0 (0%)]
364 (2%) came back with some kind of an error. [364 (2%) / 0 (0%) / 0 (0%)]
6010 (26%) have returned a successful result. [6010 (26%) / 0 (0%) / 0 (0%)]
8852 (38%) are still in progress. [8852 (38%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1448 (24%) are pending validation. [1448 (24%) / 0 (0%) / 0 (0%)]
4562 (76%) have been successfully validated. [4562 (76%) / 0 (0%) / 0 (0%)]
0 (0%) were invalid. [0 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25690066. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 4.41% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas? | |
|
|
|
|
Why is the abortion-rate that high?
Two examples found
1) all 1384 | In progress 369 | Aborted 978
2) all 63 | In progress 2 | Aborted 5
There is a program to eliminate verification tasks? | |
|
|
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
Maybe the participants tested several configurations to run the tasks (1 using all cores, 2 using half cores each, etc.) and they aborted them after estimating the different timings and keeping the best one. At least, I did so. | |
|
MonkeydeeVolunteer tester
Send message
Joined: 8 Dec 13
Posts: 467
ID: 284516
Credit: 746,846,783
RAC: 773,564
    
|
|
Another possibility is people "cherry picking" TSC servers. Start 50 instances, let the tasks run for a while, abort the tasks and power down the slowest ones and replace those with, hopefully, faster instances.
____________
My Primes
Badge Score: 2*1 + 4*2 + 6*1 + 7*10 + 8*2 + 10*3 + 11*1 = 143
 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2926
ID: 130544
Credit: 1,527,307,352
RAC: 3,772,924
|
A DC ONLY switch ...
Let me help out with the word you're having trouble with...
impossible [ im-pos-uh-buhl ]
Observe Discord where I beg to differ... | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13648
ID: 53948
Credit: 285,351,178
RAC: 47,109
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
Well, you got me interested, so I took a look.
There's a few people with a *LOT* of aborted tasks. I dug deeper and... it's not some evil strategy to dominate the world. It's not actually anything interesting at all. It's mostly a single computer which is instantly aborting *all* of its tasks. When it only takes a few seconds per task, a single computer can break many thousands per day.
My guess is the clock on the computer is set wrong; specifically, it's set to sometime in the future beyond any of the tasks' deadlines. When this happens (and we've seen this before), as soon as BOINC downloads a tasks, it realizes (erroneously) that the tasks are past their deadline, and have not yet started to run. BOINC then immediately aborts the task. Rinse and repeat, ad infinitum.
This computer probably isn't even set up to run the challenge as it's doing this on many of the sub-projects. It's doing more PSP than others because the server is set to favor challenge tasks (this improves performance.)
So there you have it. The big number of aborts is mostly due to a single broken computer.
____________
My lucky number is 75898524288+1 | |
|
dukebgVolunteer tester
Send message
Joined: 21 Nov 17
Posts: 241
ID: 950482
Credit: 23,670,125
RAC: 0
              
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
There are many scenarios in which you download more tasks than you actually want to run and abort some of them – maybe by some logic. Some of them were outlined by others above, but I'ld like to point out one more.
There's a bug in BOINC where – if you don't have any tasks for some subproject in your client_state.xml or whatever and you start running that subproject, it will download one task per each core (as if you were going to run one task per core), independent of the settings of how many cores per task you've set for the subproject. So it's very possible to get way too many tasks downloaded at the start of the challenge like this and abort most of them.
Maybe this was fixed or alleviated by the changes with how you can set multi-threading from the server side preferences now, I wouldn't know for sure (haven't crunched in... years now). | |
|
|
|
(...)
This computer probably isn't even set up to run the challenge as it's doing this on many of the sub-projects. It's doing more PSP than others because the server is set to favor challenge tasks (this improves performance.)
So there you have it. The big number of aborts is mostly due to a single broken computer.
Interesting - and good explanation which with lacking insight i could not deduce.
by the way: I changed my E5-2690 out and my E5-2670 v3 in, now 2 tasks were started with AVX and are completed with FMA3, is this a game breaker? Should i abort the tasks, or would they come out just fine?
(since nobody asked - i swapped the ssd, since the CPUs won't fit in the same board. ;) ) | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 920
ID: 301928
Credit: 513,286,703
RAC: 5,278
|
by the way: I changed my E5-2690 out and my E5-2670 v3 in, now 2 tasks were started with AVX and are completed with FMA3, is this a game breaker? Should i abort the tasks, or would they come out just fine?
Should be no problems, checkpoints are independent from transform and everything should be compatible.
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13648
ID: 53948
Credit: 285,351,178
RAC: 47,109
|
(...)
This computer probably isn't even set up to run the challenge as it's doing this on many of the sub-projects. It's doing more PSP than others because the server is set to favor challenge tasks (this improves performance.)
So there you have it. The big number of aborts is mostly due to a single broken computer.
Interesting - and good explanation which with lacking insight i could not deduce.
by the way: I changed my E5-2690 out and my E5-2670 v3 in, now 2 tasks were started with AVX and are completed with FMA3, is this a game breaker? Should i abort the tasks, or would they come out just fine?
(since nobody asked - i swapped the ssd, since the CPUs won't fit in the same board. ;) )
2 x 2 = 4 no matter whether you use AVX or FMA3 to do the multiply, so you're fine. The transform is used to get from step N to step N+1, but the result is the same no matter the method you use to do the math. The checkpoints are done between completed steps.
You can also change the number of threads between saves, although it's a bit tricky to get both BOINC and LLR to recognize the change.
That being said, completing the calculation correctly isn't the only thing you need to worry about. if you switched the CPU (and the motherboard too???) the BOINC server may (or may not) decide this is a brand new host. If it does, it will reject the tasks when they're returned. I suggest hitting the "Update" button, and see A) if there's any error messages in the log, and B) whether you see the computer with a new host ID on the PrimeGrid website.
If it created a new host, then yes, abort the tasks. If it didn't, it should be ok to continue running them.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
I only took the disk and put it in the other rig.
Done this time and again with different discs, boinc never created new hosts.
So i presume i am fine. | |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
|
I discovered something 'fun'
And I noticed this before, but I took images this time so I could remember correctly.
I have been running 16 thread tasks on a 16 core computer with no GPU tasks.
Hyper threading is turned off in the bios.
I had some 16 CPU tasks ready to go.
I changed settings and downloaded some 8 CPU tasks.
Running 2 x 8 thread tasks looked to be slower so back to the original plan.
There can be 'issues' when changing the number of cores used total, and changing the number of threads in a task. You can see the number of CPUs supposedly running (according to the tasks) as more than the number of threads available.
This took the cake - haha it was on my computer called 'Cake' too.
2 tasks running each with 16 CPUs - obviously this couldn't be true
I aborted those.
2 tasks running each with 16 CPUs
All my 16 CPU tasks had been contaminated.
I aborted them all.
Downloaded new ones.
Back to 1 task running 16 CPUs.
Didn't have the same issue to this degree with another 16 core computer.
The difference being that I had been running 15 thread tasks, and changed settings to run 2 x 8 thread tasks. When I changed settings back I still had 15 CPU tasks as only 1 task was running at a time. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 5:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-18 07:03:29 UTC)
27552 tasks have been sent out. [CPU/GPU/anonymous_platform: 27552 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
9145 (33%) were aborted. [9145 (33%) / 0 (0%) / 0 (0%)]
462 (2%) came back with some kind of an error. [462 (2%) / 0 (0%) / 0 (0%)]
8691 (32%) have returned a successful result. [8691 (32%) / 0 (0%) / 0 (0%)]
9254 (34%) are still in progress. [9254 (34%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1745 (20%) are pending validation. [1745 (20%) / 0 (0%) / 0 (0%)]
6946 (80%) have been successfully validated. [6946 (80%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25866992. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 5.13% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
GDBSend message
Joined: 15 Nov 11
Posts: 261
ID: 119185
Credit: 2,859,217,799
RAC: 357,194
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
There are many scenarios in which you download more tasks than you actually want to run and abort some of them – maybe by some logic. Some of them were outlined by others above, but I'ld like to point out one more.
There's a bug in BOINC where – if you don't have any tasks for some subproject in your client_state.xml or whatever and you start running that subproject, it will download one task per each core (as if you were going to run one task per core), independent of the settings of how many cores per task you've set for the subproject. So it's very possible to get way too many tasks downloaded at the start of the challenge like this and abort most of them.
Maybe this was fixed or alleviated by the changes with how you can set multi-threading from the server side preferences now, I wouldn't know for sure (haven't crunched in... years now).
Very likely the high abort rate is from DC task hunting. With the long running llrPSP tasks,
you may have a large amount of time left after running your last long task that will finish
before the challenge ends. So you can run DC tasks to use the remaining challenge time.
You do this by downloading 10 days of tasks, aborting all the long tasks, and keeping the
DC tasks to do after your last long task finishes
____________
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13648
ID: 53948
Credit: 285,351,178
RAC: 47,109
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
There are many scenarios in which you download more tasks than you actually want to run and abort some of them – maybe by some logic. Some of them were outlined by others above, but I'ld like to point out one more.
There's a bug in BOINC where – if you don't have any tasks for some subproject in your client_state.xml or whatever and you start running that subproject, it will download one task per each core (as if you were going to run one task per core), independent of the settings of how many cores per task you've set for the subproject. So it's very possible to get way too many tasks downloaded at the start of the challenge like this and abort most of them.
Maybe this was fixed or alleviated by the changes with how you can set multi-threading from the server side preferences now, I wouldn't know for sure (haven't crunched in... years now).
Very likely the high abort rate is from DC task hunting. With the long running llrPSP tasks,
you may have a large amount of time left after running your last long task that will finish
before the challenge ends. So you can run DC tasks to use the remaining challenge time.
You do this by downloading 10 days of tasks, aborting all the long tasks, and keeping the
DC tasks to do after your last long task finishes
While that's indeed a plausible and reasonable explanation, the actual explanation is that it's mostly due to a single broken computer -- most likely with its system date set in the future -- that is continuously and automatically aborting all tasks as soon as they're received. It's certainly not intentional.
____________
My lucky number is 75898524288+1 | |
|
GDBSend message
Joined: 15 Nov 11
Posts: 261
ID: 119185
Credit: 2,859,217,799
RAC: 357,194
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
There are many scenarios in which you download more tasks than you actually want to run and abort some of them – maybe by some logic. Some of them were outlined by others above, but I'ld like to point out one more.
There's a bug in BOINC where – if you don't have any tasks for some subproject in your client_state.xml or whatever and you start running that subproject, it will download one task per each core (as if you were going to run one task per core), independent of the settings of how many cores per task you've set for the subproject. So it's very possible to get way too many tasks downloaded at the start of the challenge like this and abort most of them.
Maybe this was fixed or alleviated by the changes with how you can set multi-threading from the server side preferences now, I wouldn't know for sure (haven't crunched in... years now).
Very likely the high abort rate is from DC task hunting. With the long running llrPSP tasks,
you may have a large amount of time left after running your last long task that will finish
before the challenge ends. So you can run DC tasks to use the remaining challenge time.
You do this by downloading 10 days of tasks, aborting all the long tasks, and keeping the
DC tasks to do after your last long task finishes
While that's indeed a plausible and reasonable explanation, the actual explanation is that it's mostly due to a single broken computer -- most likely with its system date set in the future -- that is continuously and automatically aborting all tasks as soon as they're received. It's certainly not intentional.
If you go to go to the Challenge leaderboard and look for users that have 5 times or more
tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
____________
| |
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2267
ID: 1178
Credit: 11,682,304,120
RAC: 13,086,991
                                      
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more
tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
Actually, that probably isn't the case in most instances because getting DC tasks is often done in clumps on machines that are not DC hunting specifically. For example, I have some machines that haven't yet downloaded a single DC task during the challenge, whereas others (and these are identical machines right down to adjacent serial numbers on the same wired switches) have downloaded clusters of 5 to 12 DC tasks that are completed before a main task is then gotten. The randomness of this server process and timing would even out in the long run, but over a finite period like a challenge it will appear otherwise.
| |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25866992. We need to get to n=25964955 to move from entry rank #12 to entry rank #11 on Caldwell's list. Should be doable.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
|
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more
tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
I don't think so.
As said, i have aborted around 10 % (now less) to reallign WUs (and that was a chore), so that thay are staggered an not end all around the same point and i am at 115 for around 1.4 million credits.
I am at position 59 and above me only one user (wscr) is sticking out with around 350 WUs @ 1.7 million credits compared to the surroundings and according to some of his hosts is or was actively doing this. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13648
ID: 53948
Credit: 285,351,178
RAC: 47,109
|
Why is the abortion-rate that high?
I mean, i too aborted around 10 %, to "realign" the WUs with some DC-tasks, but that high a percentage?
Any ideas?
There are many scenarios in which you download more tasks than you actually want to run and abort some of them – maybe by some logic. Some of them were outlined by others above, but I'ld like to point out one more.
There's a bug in BOINC where – if you don't have any tasks for some subproject in your client_state.xml or whatever and you start running that subproject, it will download one task per each core (as if you were going to run one task per core), independent of the settings of how many cores per task you've set for the subproject. So it's very possible to get way too many tasks downloaded at the start of the challenge like this and abort most of them.
Maybe this was fixed or alleviated by the changes with how you can set multi-threading from the server side preferences now, I wouldn't know for sure (haven't crunched in... years now).
Very likely the high abort rate is from DC task hunting. With the long running llrPSP tasks,
you may have a large amount of time left after running your last long task that will finish
before the challenge ends. So you can run DC tasks to use the remaining challenge time.
You do this by downloading 10 days of tasks, aborting all the long tasks, and keeping the
DC tasks to do after your last long task finishes
While that's indeed a plausible and reasonable explanation, the actual explanation is that it's mostly due to a single broken computer -- most likely with its system date set in the future -- that is continuously and automatically aborting all tasks as soon as they're received. It's certainly not intentional.
If you go to go to the Challenge leaderboard and look for users that have 5 times or more
tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
It's certainly possible that people are doing this, but they're not the cause of the large number of aborted tasks. If this is happening, it's only causing a smaller portion of the aborts. By a significant margin, the largest source of the aborts is a single malfunctioning computer.
What you're suggesting isn't at all unreasonable. I suspected the same thing, which is part of the reason I looked into what was happening. But this wasn't, in fact, the cause of the aborts.
____________
My lucky number is 75898524288+1 | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 920
ID: 301928
Credit: 513,286,703
RAC: 5,278
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more
tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
Actually, that probably isn't the case in most instances because getting DC tasks is often done in clumps on machines that are not DC hunting specifically. For example, I have some machines that haven't yet downloaded a single DC task during the challenge, whereas others (and these are identical machines right down to adjacent serial numbers on the same wired switches) have downloaded clusters of 5 to 12 DC tasks that are completed before a main task is then gotten. The randomness of this server process and timing would even out in the long run, but over a finite period like a challenge it will appear otherwise.
DC tasks have higher priority then normal, so a lucky computer with reasonable buffers, which connected immediately after next validation / work loading step, will often get all DC tasks loaded on that step.
| |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 6!
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-19 06:52:18 UTC)
30783 tasks have been sent out. [CPU/GPU/anonymous_platform: 30783 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
10339 (34%) were aborted. [10339 (34%) / 0 (0%) / 0 (0%)]
529 (2%) came back with some kind of an error. [529 (2%) / 0 (0%) / 0 (0%)]
11020 (36%) have returned a successful result. [11020 (36%) / 0 (0%) / 0 (0%)]
8895 (29%) are still in progress. [8895 (29%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1682 (15%) are pending validation. [1682 (15%) / 0 (0%) / 0 (0%)]
9338 (85%) have been successfully validated. [9338 (85%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=26006261. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 5.69% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
DC tasks are also a lot lower in credit. It takes ~130 DC tasks to get the same credit as a main task. Plus with these large tasks, how can there be enough DC tasks to keep up with the demand this type of hunting. It would have to be done 24/7 and the guarantee to get DC tasks with the amount of Main tasks out there would never have enough credit. Yes the number of tasks would be large, but the credit would be small. If you look at the last update 30K+ has been sent out, that counts both Main and DC tasks. Yes 10K+ has been aborted, but that means those 10K do not have DC tasks yet, so only ~20K are live, of that maybe half are DC. 10K DC tasks are not going to put someone on the leaderboard, and as has been proven by admins of the system, DC tasks are going to other people. I have had many myself. I also have older tasks still waiting for their DC to return, which does not look like they would have gone to DC miners.
Think about the logistics before making assumptions.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1 | |
|
PDWSend message
Joined: 14 Nov 14
Posts: 21
ID: 373199
Credit: 1,730,343,278
RAC: 731
      
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
DC tasks are also a lot lower in credit. It takes ~130 DC tasks to get the same credit as a main task. Plus with these large tasks, how can there be enough DC tasks to keep up with the demand this type of hunting. It would have to be done 24/7 and the guarantee to get DC tasks with the amount of Main tasks out there would never have enough credit. Yes the number of tasks would be large, but the credit would be small. If you look at the last update 30K+ has been sent out, that counts both Main and DC tasks. Yes 10K+ has been aborted, but that means those 10K do not have DC tasks yet, so only ~20K are live, of that maybe half are DC. 10K DC tasks are not going to put someone on the leaderboard, and as has been proven by admins of the system, DC tasks are going to other people. I have had many myself. I also have older tasks still waiting for their DC to return, which does not look like they would have gone to DC miners.
Think about the logistics before making assumptions.
As Michael has stated in each update including the last one...
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-19 06:52:18 UTC)
30783 tasks have been sent out. | |
|
GDBSend message
Joined: 15 Nov 11
Posts: 261
ID: 119185
Credit: 2,859,217,799
RAC: 357,194
|
If you go to go to the Challenge leaderboard and look for users that have 5 times or more tasks than their neighbors, they're probably DC task hunting, and aborting a lot of tasks.
DC tasks are also a lot lower in credit. It takes ~130 DC tasks to get the same credit as a main task. Plus with these large tasks, how can there be enough DC tasks to keep up with the demand this type of hunting. It would have to be done 24/7 and the guarantee to get DC tasks with the amount of Main tasks out there would never have enough credit. Yes the number of tasks would be large, but the credit would be small. If you look at the last update 30K+ has been sent out, that counts both Main and DC tasks. Yes 10K+ has been aborted, but that means those 10K do not have DC tasks yet, so only ~20K are live, of that maybe half are DC. 10K DC tasks are not going to put someone on the leaderboard, and as has been proven by admins of the system, DC tasks are going to other people. I have had many myself. I also have older tasks still waiting for their DC to return, which does not look like they would have gone to DC miners.
Think about the logistics before making assumptions.
It's not that we WANT to do DC tasks. It's that we HAVE to do DC tasks. I have a laptop that
takes 3.5 days per long llrPSP task. I can finish 2 long tasks within the challenge, but a third
long task would finish after the challenge ends and get zero credit. To not waste 3 days of
challenge time, I set my download queue to 10 days, and kept downloading and aborting
long tasks until I had about 3 days of DC tasks (130). It took me about 10 minutes 2 days ago.
____________
| |
|
tngSend message
Joined: 29 Aug 10
Posts: 426
ID: 66603
Credit: 32,363,290,508
RAC: 31,817,446
                                
|
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=25866992. We need to get to n=25964955 to move from entry rank #12 to entry rank #11 on Caldwell's list. Should be doable.
Done.
But can we make #10? Redouble your efforts!
____________
 | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 7.
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-20 06:34:51 UTC)
34590 tasks have been sent out. [CPU/GPU/anonymous_platform: 34590 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
11622 (34%) were aborted. [11622 (34%) / 0 (0%) / 0 (0%)]
591 (2%) came back with some kind of an error. [591 (2%) / 0 (0%) / 0 (0%)]
13288 (38%) have returned a successful result. [13288 (38%) / 0 (0%) / 0 (0%)]
9089 (26%) are still in progress. [9089 (26%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
2014 (15%) are pending validation. [2014 (15%) / 0 (0%) / 0 (0%)]
11274 (85%) have been successfully validated. [11274 (85%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=26177853. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 6.39% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
Another reason for a high number of aborts might be trouble with Boinc client 7.16.xx.
On my Ryzen with this exact client I got several months worth of SoB tasks during the last challenge. Crunched SoB for several weeks after the challenge, then aborted the rest.
I'm running PSP in two vbox instances with Boinc client 7.14.xx now, so I won't encounter similar problems.
____________
Greetings, Jens
1461*2^3373383+1 | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 8:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-21 05:51:42 UTC)
37697 tasks have been sent out. [CPU/GPU/anonymous_platform: 37697 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
12516 (33%) were aborted. [12516 (33%) / 0 (0%) / 0 (0%)]
987 (3%) came back with some kind of an error. [987 (3%) / 0 (0%) / 0 (0%)]
15371 (41%) have returned a successful result. [15371 (41%) / 0 (0%) / 0 (0%)]
8823 (23%) are still in progress. [8823 (23%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
2199 (14%) are pending validation. [2199 (14%) / 0 (0%) / 0 (0%)]
13172 (86%) have been successfully validated. [13172 (86%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=26316023. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 6.95% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Day 9:
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-22 04:28:16 UTC)
43484 tasks have been sent out. [CPU/GPU/anonymous_platform: 43484 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
15058 (35%) were aborted. [15058 (35%) / 0 (0%) / 0 (0%)]
1019 (2%) came back with some kind of an error. [1019 (2%) / 0 (0%) / 0 (0%)]
17148 (39%) have returned a successful result. [17148 (39%) / 0 (0%) / 0 (0%)]
10259 (24%) are still in progress. [10259 (24%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
2516 (15%) are pending validation. [2516 (15%) / 0 (0%) / 0 (0%)]
14632 (85%) have been successfully validated. [14632 (85%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=26535147. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 7.84% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
A few hours left in the challenge! Some friendly reminders... :)
At the Conclusion of the Challenge
When the challenge completes, we would prefer users "moving on" to finish those tasks they have downloaded, if not then please ABORT the WU's (and then UPDATE the PrimeGrid project) instead of DETACHING, RESETTING, or PAUSING.
ABORTING WU's allows them to be recycled immediately; thus a much faster "clean up" to the end of a Challenge. DETACHING, RESETTING, and PAUSING WU's causes them to remain in limbo until they EXPIRE. Therefore, we must wait until WU's expire to send them out to be completed.
Likewise, if you're shutting down the computer for an extended period of time, or deleting the VM (Virtual Machine), please ABORT all remaining tasks first. Also, be aware that merely shutting off a cloud server doesn't stop the billing. You have to destroy/delete the server if you don't want to continue to be charged for it.
Thank you!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
For this challenge I turned SMT off in my BIOS following your suggestions (and it really made a difference!)
Now that the challenge is about to finish and I'll change preferences to download tasks for other subprojects, what do you recommend? Should I let SMT inactive? Are there other applications that prefer SMT on? | |
|
|
|
|
arakelov wrote: For this challenge I turned SMT off in my BIOS following your suggestions (and it really made a difference!)
There is an alternative: Leave SMT on but let BOINC use only half of the logical CPUs. You have got a Linux computer; the process scheduler of the Linux kernel pursues a policy of spreading the total workload across all physical cores, generally.
arakelov wrote: Now that the challenge is about to finish and I'll change preferences to download tasks for other subprojects, what do you recommend? Should I let SMT inactive? Are there other applications that prefer SMT on?
The currently active PrimeGrid subprojects which are worthwhile to run on the CPU (instead of the GPU) all look to me as if they are similar in the regard that some might have a very small throughput benefit from SMT, while in other cases SMT decreases throughput. However, I am not sure about the Ryzen 7 3700X in particular. However, even in cases when the currently active PrimeGrid subprojects benefit from SMT with respect to throughput, there will always be a regression in electric energy spent on the same amount of work (expressed in Joule per credit, or in Watts per points-per-day). Hence it is generally preferable not to use SMT when running PrimeGrid CPU-only applications, in my experience. | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
|
The CPU runs considerably cooler on my system (and I guess more efficiently) when you disable the logical cores. Under Linux that's easy and done on-the-fly:
echo 0 > /sys/devices/system/cpu/cpu0/online
Instead of cpu0 use the core #, e.g. cpu1, cpu2 and so on. Usually the second half of the cores are the logical ones. Thus, for a 10-core system: cpu10 to cpu19. Or check via cat /proc/cpuinfo which processors have the same core id.
It should also be possible to use
echo off > /sys/devices/system/cpu/smt/control
to disable SMT. I would assume the scheduler will shuffle the threads from core to core which won't happen if you put them offline.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
|
|
|
Bur wrote: I would assume the scheduler will shuffle the threads from core to core
No, why would it? At least on Linux, the scheduler lets a software thread remain on the same logical CPU as long as feasible, because that's a more efficient use of the caches than kicking a thread unnecessarily around different logical CPUs. Furthermore, the scheduler is aware of HT/SMT hardware thread siblings among the logical CPUs and attempts to spread the overall system load across all physical cores. (This is from memory and aligns with what I observe, alas I don't have links to sources.) | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
|
I was expecting that, but when running cado factoring software that uses 5 processes with 2 threads each, the load on the 20 cores (10 physical) was switching around all the time. Maybe it was just moving within the same physical core, but it still seemed unneccessary.
But of course, I never ran any tests or did research. It was just this one piece of software and I only had a look at htop output.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
|
|
|
The answer to the posts of you both is: Your mileage may vary.
You could sometimes gain performance with SMT/HT enabled, you could sometimes lose performance.
It is not that deterministic and depends on the workload and CPU-architecture - so there is no "one answer fits all"-guideline.
Normally i would disable SMT/HT out of security reasons - but on my Thinkpad A285 with Ryzen 7 Pro 2700U for instance i found a bug with the nosmt-implementation where it draws more power and runs way hotter (75 °C vs 50 °C) with nosmt versus smt enabled, so i switched back. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
And we're done! Phenomenal work to everyone, we continue to blow previous records out of the water. Cleanup tasks now available, final stats below!
Challenge: Once In a Blue Moon
App: 8 (PSP-LLR)
Fast DC tasks are NOT included.
(As of 2021-08-23 06:32:50 UTC)
51121 tasks have been sent out. [CPU/GPU/anonymous_platform: 51121 (100%) / 0 (0%) / 0 (0%)]
Of those tasks that have been sent out:
24138 (47%) were aborted. [24138 (47%) / 0 (0%) / 0 (0%)]
1037 (2%) came back with some kind of an error. [1037 (2%) / 0 (0%) / 0 (0%)]
18524 (36%) have returned a successful result. [18524 (36%) / 0 (0%) / 0 (0%)]
6042 (12%) are still in progress. [6042 (12%) / 0 (0%) / 0 (0%)]
Of the tasks that have been returned successfully:
1926 (10%) are pending validation. [1926 (10%) / 0 (0%) / 0 (0%)]
16598 (90%) have been successfully validated. [16598 (90%) / 0 (0%) / 0 (0%)]
2 (0%) were invalid. [2 (0%) / 0 (0%) / 0 (0%)]
0 (0%) are inconclusive. [0 (0%) / 0 (0%) / 0 (0%)]
The current leading edge (i.e., latest work unit for which work has actually been sent out to a host) is n=26615093. The leading edge was at n=24605319 at the beginning of the challenge. Since the challenge started, the leading edge has advanced 8.17% as much as it had prior to the challenge!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
|
|
|
Thank you to those who helped organise this focused event on the PSP subproject.
I at least received a new badge out of it.
____________
| |
|
Nick
Send message
Joined: 11 Jul 11
Posts: 1006
ID: 105020
Credit: 2,734,196,843
RAC: 9,126,537
|
|
I aborted 2.35% of those while doing 0.89% of the work.
Why?
For some unknown, there are 2 speeds of tasks with PSP and SOB that is not FFT.
There is normal speed and about 1.5 times longer.
It may be an AVX512 thing.
It makes no sense!
(Chewbaccer defence) | |
|
compositeVolunteer tester
Send message
Joined: 16 Feb 10
Posts: 841
ID: 55391
Credit: 784,035,755
RAC: 431,049
     
|
I only had a look at htop output.
1-second sampling htop doesn't gives you the whole picture.
You can adjust the sampling rate with a command line flag to get a better idea.
You typically won't see fleeting tasks like interrupt handlers running, which will take over a logical core and kick off whatever is running there for less time than it takes to render the window. That's why tasks randomly switch logical cores. If the scheduler is smart enough, it should realize that tasks with lots of cache still loaded on a particular core would be better off returning to the same core instead of trying to run the task earlier on another core and causing a spike in RAM accesses. It depends on how busy the system is. | |
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 462
ID: 1241833
Credit: 279,915,382
RAC: 551,580
|
|
The 10 threads were running at 100% load, so I could see how the ten 100% loads moved from core to core. E.g. in one second core 6 was idle, the next second it had 100% load, then idle again. But in total it was always 10 fully loaded cores just not the same ones. My observation is just based on one application, though, so I'm not saying the scheduler is no good. I just disabled the cores because in my usage scenario (system only used for one task that completely utilizes all cores) they don't make sense, I think.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,800,000 | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
Cleanup Status:
Aug 22: 2399 tasks outstanding; 1677 affecting individual (232) scoring positions; 807 affecting team (50) scoring positions.
Aug 23: 890 tasks outstanding; 556 affecting individual (152) scoring positions; 234 affecting team (24) scoring positions.
Aug 24: 541 tasks outstanding; 303 affecting individual (117) scoring positions; 37 affecting team (14) scoring positions.
Aug 25: 371 tasks outstanding; 190 affecting individual (91) scoring positions; 22 affecting team (10) scoring positions.
Aug 26: 318 tasks outstanding; 128 affecting individual (77) scoring positions; 15 affecting team (8) scoring positions.
Aug 27: 200 tasks outstanding; 77 affecting individual (50) scoring positions; 9 affecting team (6) scoring positions.
Aug 28: 79 tasks outstanding; 26 affecting individual (20) scoring positions; 3 affecting team (3) scoring positions.
Aug 29: 59 tasks outstanding; 22 affecting individual (17) scoring positions; 3 affecting team (3) scoring positions.
Aug 30: 34 tasks outstanding; 10 affecting individual (9) scoring positions; 3 affecting team (3) scoring positions.
Aug 31: 21 tasks outstanding; 7 affecting individual (7) scoring positions; 3 affecting team (3) scoring positions.
Sep 1: 6 tasks outstanding; 4 affecting individual (4) scoring positions; 2 affecting team (2) scoring positions.
Sep 2: 4 tasks outstanding; 3 affecting individual (3) scoring positions; 2 affecting team (2) scoring positions.
Sep 3: 4 tasks outstanding; 3 affecting individual (3) scoring positions; 2 affecting team (2) scoring positions.
Sep 4: 4 tasks outstanding; 3 affecting individual (3) scoring positions; 2 affecting team (2) scoring positions.
Sep 5: 3 tasks outstanding; 2 affecting individual (2) scoring positions; 1 affecting team (1) scoring positions.
Sep 6: 3 tasks outstanding; 2 affecting individual (2) scoring positions; 1 affecting team (1) scoring positions.
Sep 7: 2 tasks outstanding; 1 affecting individual (1) scoring positions; 1 affecting team (1) scoring positions.
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
Michael GutierrezVolunteer moderator
Project administrator
Project scientist
Send message
Joined: 21 Mar 17
Posts: 189
ID: 764476
Credit: 42,875,942
RAC: 9,142
|
|
The results are final!
During these 10 days, we completed 18,524. Unfortunately the blue moon did not bring us the luck of a new prime, but we still did a phenomenal amount of work, since each task is so large. The challenge boosted the project's average output to 20 times typical.
119 teams and 502 individuals participated in the Challenge.
Top Three Individuals:
1. vaughan
2. Skillz
3. tng
Top Three Teams:
1. TeAm AnandTech
2. Antarctic Crunchers
3. SETI.Germany
The next Challenge is Martin Gardner's Birthday Challenge, which starts October 21. We hope to see you there! Thank you again to everyone. We really appreciate your *phenomenal* participation and hope you had fun!
____________
Time flies like an arrow. Fruit flies like a banana. | |
|
Post to thread
Message boards :
Number crunching :
Once in a Blue Moon Challenge |
